Algorithmic Week In Review: 2022-08-07
I’d like to use this report to look back more generally on last week, and to look forward - both at how we expect our current algorithms to perform and what we have in the works. Without further ado, let’s dive in.
Current Algorithms - Recent Performance
This week was a tough one for us, but largely for reasons we can pinpoint. If you’ve been keeping up with daily algorithm reviews, this isn’t new information. On days the market has won, we’ve underperformed it. On days the market has lost, we’ve mostly lost with it.
In general, we’ve underperformed the market because we’re playing conservatively. I went into more depth on this on the 2022-08-03 blog, but to summarize: we don’t think the market is done coming down. It’s rebounded well since hitting its bottom this year (up about 15% from a 25% drop from its high), but signs we’re looking at don’t bode well. We don’t know exactly when we’re forecasting this follow-up drop, so we’re sticking with safer bets in the meantime.
This overlay of the S&P during the 2008 crash and right now illustrates our fears nicely
As for when the market was down, we mostly underperformed due to individual stocks having bad news unique to them (for example, $WMT in the 2022-08-04 report). Our algorithms aren’t able to account for news - only technical signals. Our hedging algorithms can also only hedge against movements by the market as a whole, or individual sectors - one stock moving down hard isn’t something they can resist.
Current Algorithms - Performance Looking Forward
I hate to say it, but I suspect this week will look a lot like last week for us. We’re still holding a defensive portfolio in an environment where, short term, growth stocks have performed well. We’re going to be sticking with this strategy for the foreseeable future - until we feel the market has well and truly bottomed out. I’m expecting the Base Algo and Neural Net to underperform the market, but remain positive (at least until that drop comes). But after the drop, Kokomo and GoodVibrations (our hedging systems) will probably have done the best. This is a tough moment for us, but we’re choosing to sacrifice short-term success for long term safety and stability.
Here are our algorithms’ exposures for Monday. No major changes here; they look fairly similar to their exposures throughout last week. As always, we like the Base Algorithm and Neural Net (personally, I prefer the base) on days we’re bullish on the market. On days we’re bearish, we lean more towards Kokomo and GoodVibrations.
Future Algorithms - What’s In The Pipeline?
I would like to look into automated trading. Getting into and out of positions is a major time sink for Allen each day, so he has requested this. It’ll also enable us to explore more complex strategies and positions. Currently, this is low priority. It won’t affect our profits until we have a strategy we like that’s too complex for Allen to do by hand every day.
I’m mostly focusing on hedging systems now, because there’s time pressure. We have to have them ready by the time the market drop we’re expecting comes. If we come up with a great new market hedge when SPY is already rebounding, so what? Among other things, I’m considering exactly how GoodVibrations assigns its sector betas to each ticker. For instance, consider $AAPL. Is this technology? Consumer Discretionary? Maybe it’s both, to their full normal levels? Or some exponential weighted average of the two?
Lastly, mixing algorithms together. Our algorithms all examine trading signals the same way they hedge their holdings (for example, when Kokomo looks at the RSI of a ticker, it looks at it controlled for influence from its market beta). But, after a weekend of testing and recombining, we can improve performance.
We have a combination of the Base Algo and GoodVibrations (Base Vibrations?) in testing right now. Tentatively, it outperforms our other systems by using raw price data to predict sector-independent returns. We aren’t firm enough on it to start publishing, but we’ll be watching it closely for the next week or so to gauge performance. If things go well, expect to see it in our beta channel, and later in our releases for all users.
That’s all for now. Thank you for reading. Good luck trading this week, everyone.
-Asher
